You replace the first label with one that sorts after it as the next name. Append a 0x00 octet if there is space otherwise increment the last octet by one. If that overflows remove the last octet and increment the second last octet. If that overflows remove it and increment the new last octet. Repeat. If the label is exhausted apply this to the second last label. If it is in the zone otherwise use the zone origin. 



-- 
Mark Andrews

On 4 Nov 2026, at 23:12, Rifat Jumagulov <[email protected]> wrote:


Hello,

RFC 4470 Section 4 builds the next name this way: "To increment a name, add a leading label with a single null (zero-value) octet." For https://www.google.com/url?q=http://foo.example.com&source=gmail&ust=1793623136948000&sa=E it gives an NSEC record with next name \https://www.google.com/url?q=http://000.foo.example.com&source=gmail&ust=1793623136948000&sa=E, and it says: "The first of these NSEC RRs proves that no exact match for https://www.google.com/url?q=http://foo.example.com&source=gmail&ust=1793623136948000&sa=E exists"

RFC 8198 Appendix B reads the same form differently: "If the next domain name is a subdomain of the given name, you have an ENT. Otherwise, you have an NXDOMAIN." \https://www.google.com/url?q=http://000.foo.example.com&source=gmail&ust=1793623136948000&sa=E is a subdomain of https://www.google.com/url?q=http://foo.example.com&source=gmail&ust=1793623136948000&sa=E, so this NSEC shows an empty non-terminal, not a missing name. Section 3 of RFC 4470 also says: "This change is expected to be fully compatible with all existing DNSSEC validators."

Erratum 6734 (Reported, Mark Andrews) makes the general point: \000.domain can only prove no data at domain, not that domain does not exist. It has no corrected text yet. RFC 7129 Appendix A prints the same form, NSEC \https://www.google.com/url?q=http://000.b.example.org&source=gmail&ust=1793623136948000&sa=E.

Local test: a zone with the two NSEC records of the Section 4 example, signed with ECDSAP256SHA256, served by NSD 4.15.4 and by Knot DNS 3.6.2, which both return NXDOMAIN with both records. Query https://www.google.com/url?q=http://foo.example.com&source=gmail&ust=1793623136948000&sa=E. A with DO=1. BIND 9.20.30, Unbound 1.24.0, Knot Resolver 6.5.1 and PowerDNS Recursor 5.4.9 all answer SERVFAIL. https://www.google.com/url?q=http://example.com&source=gmail&ust=1793623136948000&sa=E. SOA gets AD from all four, so the key and anchor work. As a control I signed the same zone with next names foo\https://www.google.com/url?q=http://000.example.com&source=gmail&ust=1793623136948000&sa=E. and *\https://www.google.com/url?q=http://000.example.com&source=gmail&ust=1793623136948000&sa=E., which are not children. All four then answer NXDOMAIN with AD.

Could the WG give erratum 6734 corrected text that stops recommending \000.QNAME as a next name for an NXDOMAIN proof?

AI models helped review the text and build small test models; Arxo ran the checks, and I take responsibility for the reports.

Rifat Jumagulov
Arxo Law · [email protected]
_______________________________________________
DNSOP mailing list -- [email protected]
To unsubscribe send an email to [email protected]
_______________________________________________
DNSOP mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to